Seven Mutations in the Human Insulin Gene Linked to Permanent Neonatal/Infancy-Onset Diabetes Mellitus

Seven mutations in the human insulin gene linked to permanent neonatal/infancy-onset diabetes mellitus

Carlo Colombo, … , Fabrizio Barbetti, the Early Onset Diabetes Study Group of the Italian Society of Pediatric Endocrinology and Diabetes (SIEDP)

J Clin Invest. 2008;118(6):2148-2156. https://doi.org/10.1172/JCI33777.

Research Article Metabolism

Permanent neonatal diabetes mellitus (PNDM) is a rare disorder usually presenting within 6 months of birth. Although several genes have been linked to this disorder, in almost half the cases documented in Italy, the genetic cause remains unknown. Because the Akita mouse bearing a mutation in the Ins2 gene exhibits PNDM associated with pancreatic β cell apoptosis, we sequenced the human insulin gene in PNDM subjects with unidentified mutations. We discovered 7 heterozygous mutations in 10 unrelated probands. In 8 of these patients, insulin secretion was detectable at diabetes onset, but rapidly declined over time. When these mutant proinsulins were expressed in HEK293 cells, we observed defects in insulin protein folding and secretion. In these experiments, expression of the mutant proinsulins was also associated with increased Grp78 protein expression and XBP1 mRNA splicing, 2 markers of endoplasmic reticulum stress, and with increased apoptosis. Similarly transfected INS-1E insulinoma cells had diminished viability compared with those expressing WT proinsulin. In conclusion, we find that mutations in the insulin gene that promote proinsulin misfolding may cause PNDM.

Find the latest version: https://jci.me/33777/pdf Research article Seven mutations in the human insulin gene linked to permanent neonatal/infancy-onset diabetes mellitus Carlo Colombo,1 Ottavia Porzio,2 Ming Liu,3 Ornella Massa,1 Mario Vasta,4 Silvana Salardi,5 Luciano Beccaria,6 Carla Monciotti,7 Sonia Toni,8 Oluf Pedersen,9 Torben Hansen,9 Luca Federici,10 Roberta Pesavento,11 Francesco Cadario,12 Giorgio Federici,2 Paolo Ghirri,13 Peter Arvan,3 Dario Iafusco,14 Fabrizio Barbetti,1,2,15 and the Early Onset Diabetes Study Group of the Italian Society of Pediatric Endocrinology and Diabetes (SIEDP)

1Laboratory of Molecular Endocrinology and Metabolism, Bambino Gesù Children’s Hospital, Scientific Institute (IRCCS), Rome, Italy. 2Department of Internal Medicine, University of Tor Vergata, Rome, Italy. 3Division of Metabolism, Endocrinology, and Diabetes, Department of Internal Medicine, University of Michigan Health System, Ann Arbor, Michigan, USA. 4Diabetes Clinic, Urbino City Hospital, Urbino, Italy. 5Department of Pediatrics, University of Bologna, Bologna, Italy. 6Department of Pediatrics, Alessandro Manzoni Hospital, Lecco, Italy. 7Department of Pediatrics, University of Padua, Padua, Italy. 8Regional Center for Juvenile Diabetes, Meyer Pediatric Hospital, Florence, Italy. 9Steno Diabetes Center, Gentofte, Denmark. 10Ce.S.I. Center for Studies on Aging, G. d’Annunzio University Foundation and Department of Biomedical Sciences, University of Chieti G. d’Annunzio, Chieti, Italy. 11Pediatric Unit, Boldrini Hospital, Thiene, Italy. 12Department of Pediatrics, University of Eastern Piedmont, Interdisciplinary Research Center of Autoimmune Diseases, Novara, Italy. 13Department of Pediatrics, Division of Neonatology, University of Pisa, Pisa, Italy. 14Department of Pediatrics, 2nd University of Naples, Naples, Italy. 15Biomedical Park Foundation S. Raffaele, Rome, Italy.

Permanent neonatal diabetes mellitus (PNDM) is a rare disorder usually presenting within 6 months of birth. Although several genes have been linked to this disorder, in almost half the cases documented in Italy, the genetic cause remains unknown. Because the Akita mouse bearing a mutation in the Ins2 gene exhibits PNDM associated with pancreatic β cell apoptosis, we sequenced the human insulin gene in PNDM subjects with unidentified mutations. We discovered 7 heterozygous mutations in 10 unrelated probands. In 8 of these patients, insulin secretion was detectable at diabetes onset, but rapidly declined over time. When these mutant proinsulins were expressed in HEK293 cells, we observed defects in insulin protein folding and secretion. In these experiments, expression of the mutant proinsulins was also associated with increased Grp78 protein expression and XBP1 mRNA splicing, 2 markers of endoplasmic reticulum stress, and with increased apoptosis. Similarly transfected INS-1E insulinoma cells had diminished viability compared with those expressing WT proinsulin. In conclusion, we find that mutations in the insulin gene that promote proinsulin misfolding may cause PNDM.

Introduction birth, but delayed onset has been reported in subjects with muta- Neonatal diabetes is a rare monogenic disorder associated with tions of EIF2AK3 (6, 14). PNDM can manifest itself as an isolated defects of pancreatic β cell mass and/or function. The disorder defect in glucose-stimulated insulin secretion (mutations of GCK, is distinguished by 2 clinical variants — transient and permanent KCNJ11, and ABCC8; refs. 4, 7, 9) or as part of a syndrome involving — caused by specific genetic defects. Abnormalities of chromo- other organ systems (4–6, 8–11, 15). Most patients (i.e., 40%–64%) some 6 or mutations of the ATP-sensitive potassium (KATP) channel with PNDM or infancy-onset diabetes bear dominant, activating, KCNJ11 and ABCC8 genes are detected in most patients with tran- heterozygous mutations of KCNJ11, the pore-forming subunit sient neonatal diabetes mellitus (TNDM), in which diabetes usually of KATP channel (15), while activating mutations of its regulatory presents in the first month of life but then remits with a variable subunit SUR1 (ABCC8 gene) account for an additional 7% of cases time interval (1–4). In about 50% of patients with TNDM, diabetes (4). Other genetic defects leading to PNDM are much less common relapses later in life (1–4). Permanent neonatal diabetes mellitus (5–8, 10, 11). We previously identified KCNJ11 mutations in 19 of (PNDM) is more genetically heterogeneous and has been associated 37 patients in an Italian cohort with diabetes onset by 7 mo of age with mutations in 8 different genes: IPF-1, EIF2AK3, GCK, FOXP3, (13, 16, 17), and we set about to search for the genetic causes of KCNJ11, PTF1A, ABCC8, and GLIS3 (4–11). Individuals with PNDM diabetes in the remaining subjects. need insulin therapy, but those carrying mutations in KCNJ11 and In one of these patients, kindred A, II-1, the C-peptide level ABCC8 can be successfully transferred to sulfonylureas (4, 12, 13). at the time of diagnosis was surprisingly high, unlike the low In most PNDM, diabetes manifests itself within the first 6 mo after C-peptide level in patients with PNDM caused by mutations of GCK, KCNJ11, and ABCC8 (4, 7, 9). High C-peptide immunoreactivity in the absence of apparent insulin resistance (18) has been iden- Nonstandard abbreviations used: hAkita, Akita mutation introduced into human proinsulin; MDI, monogenic diabetes of infancy; PNDM, permanent neonatal diabe- tified in adult patients bearing point mutations in the proinsulin tes mellitus; TNDM, transient neonatal diabetes mellitus. coding sequence leading to familial hyperproinsulinemia (19–21). Conflict of interest: The authors have declared that no conflict of interest exists. Interestingly, a heterozygous mutation in the insulin-2 gene (Ins2) Citation for this article: J. Clin. Invest. 118:2148–2156 (2008). doi:10.1172/JCI33777. has previously been identified as the cause of early-onset diabetes

2148 The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 research article

Figure 1 Family trees of patients carrying INS gene mutations. Pherograms of the mutations are also shown, with mutations indicated by black arrows. The mutation causing diabetes in kindred D, LB15YB16delinsH, was determined by subcloning of the PCR product and DNA sequencing, confirmed in 10 clones. Mutations LB11P and YA19X, which respectively introduce HpaII and SpeI restriction sites, were both confirmed (white arrows) by RFLP-PCR. n, normal allele; m, mutant allele; p, proband. Patients carrying mutations are denoted by filled symbols.

in the Akita mouse (22), in which a Cys→Tyr substitution at proin- according to their position in the mature insulin chains: LB6P (that sulin position 96 (C96Y; encoding residue 7 of the insulin A-chain) is, a Leu→Pro substitution at residue 6 of the insulin B-chain), disrupts one of the 2 disulfide bonds connecting the insulin B- LB6V, LB11P, LB15YB16delinsH (that is, replacement of Leu and Tyr and A-chains. As a consequence, the Akita proinsulin accumulates by His at residues 15–16 of the B-chain), CA6Y, and YA19X. The proin the ER, inducing ER stress and apoptosis of pancreatic β cells insulin mutation located in the dibasic doublet between C-peptide (22–25). We hypothesized that certain mutations of the human and A-chain (K64–R65, based on the N-terminal Phe residue of the insulin gene (INS) might also give rise to neonatal diabetes mel- insulin B-chain as residue 1) is denoted as R65C. In all but 2 fami- litus. In this report we describe 7 proinsulin mutations, which we lies, the mutation originated de novo, and neonatal diabetes was believe to be novel, associated with PNDM or infancy-onset diabe- observed only in subjects carrying the INS gene mutation and not tes mellitus that are likely to promote proinsulin misfolding, ER in unaffected parents. Mutation R65C was identified as a sponta- stress, and β cell failure. neous mutation as well as in 2 familial cases (Figure 1). Mutation CA6Y was also detected in a Danish patient (diabetes onset, age 3 Results mo); however, DNA samples from 2 children of the proband, who Mutational analysis. We identified 7 heterozygous mutations of the presented with diabetes at 6.5 and 2.5 mo postpartum, were not INS gene in 10 unrelated probands with PNDM or infancy-onset available for analysis. All mutations found in patients with PNDM diabetes. As would be expected for a defective gene specific to or infancy-onset diabetes were different from those previously β cells, all affected neonates had only diabetes and no other clini- described to lead to familial hyperinsulinemia or hyperproinsu- cal features. We adopted a nomenclature describing the mutations linemia (19–21, 26–32). Notably, all affected residues — including

The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 2149 research article

Table 1 Clinical characteristics of patients

Family subject Kindred A Kindred B Kindred C Kindred D Kindred E Kindred F Kindred G Kindred H Kindred I Proband Mother Proband Proband Father Proband Proband Proband Proband Proband Proband (II-1) (I-2) (II-1) (II-1) (I-1) (II-1) (II-1) (II-1) (II-3) (II-2) (II-2) Mutation R65C R65C R65C R65C R65C LB15YB16delinsH LB11P CA6Y YA19X LB6P LB6V At presentation Age 70 dA 1 yr 134 d 86 d 4 yr 37 d 69 d 75 d 88 d 139 dA 184 d Presentation HG NA KA KA HG KA HG KA HG HG KA Glucose (mg/dl) 234 NA 612 521 216 594 610 >1,000 504 306 702 C-peptide (ng/ml)B 8.5 (78 d) NA 0.85 (190 d) 0.5 (90 d) NA 0.57 (40 d) 0.7 (71 d) 0.97 (88 d) NA 3.2 (148 d) 0.8 (226 d) 3.8 (87 d) 0.14 (195 d) 4.17C (194 d) 1.17 (150 d) At termination of the present study Age 12 yr 41 yr 6 yr 16 mo 37 yr 13 yr 21 yr 28 mo 13 yr 28 mo 30 mo Insulin dose (U/kg/d) 1.41 0.67 0.6 0.37 0.83 1 1 0.6 1.14 0.38 0.85 C-peptide (ng/ml) ND NA 0.84 ND NA ND ND ND ND ND 0.8

HG, hyperglycemia; KA, ketoacidosis; NA, not available; ND, not detectable. ADiscovery of diabetes before symptoms. BNumbers in parentheses denote age at which value was determined. CAssayed while on 0.72 mg/kg/d glibenclamide with no insulin therapy. To convert to SI units, multiply glucose by 0.0555 to get mmol/l and multiply C-peptide by 331 to get pmol/l.

CA6 (and CA11), LB6, LB11, and dibasic amino acids RR and KR, at the (kindred A, I-1, 30 ng/dl at 38 mg/dl plasma glucose; kindred D, junctions between B-chain and C-peptide and between C-peptide 22 ng/dl at 34 mg/dl plasma glucose). and A-chain, respectively — are conserved among human, cow, pig, Clinical and metabolic features of PNDM or infancy-onset diabetes dog, rat, and mouse insulin I and II; between frog insulin I and II; caused by INS gene mutations compared with those from other causes. In and between eel and hagfish (33) (Supplemental Figure 1; supple- all but 1 patient with an INS gene mutation (kindred F; Table 1), mental material available online with this article; doi:10.1172/ birth weight was near normal (median, 3,055 ± 227 g). This find- JCI33777DS1). We did not detect these mutations in 200 Italian ing was in sharp contrast with the birth weight of patients with subjects with normal glucose tolerance. KCNJ11 mutations (median, 2,455 ± 370 g; P < 0.00021, INS versus Clinical characteristics. All patients were the product of uneventful KCNJ11). At the time of diagnosis, C-peptide was barely detect- pregnancies; their clinical features at presentation of diabetes are able in 4 of 10 KCNJ11 mutants of the Italian cohort (refs. 13, 16, shown in Table 1. In 3 subjects (kindreds E, G, and H), failure to and F. Barbetti, unpublished observations), whereas in patients thrive or intercurrent infection resulted in further analysis lead- for whom C-peptide data were available, measurable or even high ing to a diagnosis of diabetes. In the case of patient II-1 of kindred C-peptide levels were found in all 8 individuals with INS muta- A (mutation R65C), glycosuria was checked because the mother tions. The diagnosis of diabetes in patients with INS mutations had been diagnosed with insulin-dependent diabetes at 1 yr of age and was found positive in the diaper at 76 d postpartum. Six patients (kindreds B, C, D, F, I, and PN-30) were referred to a local hospital for severe hyperglycemia with ketosis. C-peptide, randomly determined 1–56 d after discov- ery of diabetes and during the course of insulin therapy, was detected in 8 probands (data from 2 probands and 2 affected parents were not available). In 2 of these 8 probands, C-peptide appeared unexpectedly high (kindred A, II-1, and kindred H; Table 1). Furthermore, sequential assay of C-peptide in 2 patients on insulin therapy (kindred A, II-1, and kindred F) and in one on sulfonylurea for 60 d followed by mixed insulin-sulfonylurea therapy (kindred H) revealed declining values over the ensuing months after initial diagnosis. At the termination of the present study, C-peptide was unde- tectable in all but 3 patients. Basal pancreatic glucagon level was found to be in the normal to high range in 2 subjects (kindred A, I-1, 21.3 ng/dl; kindred D, 16.5 Figure 2 ng/dl), and glucagon secretion, as determined by an Amino acid sequences of WT and mutant proinsulins. The amino acid sequence insulin tolerance test, increased approximately 50% as a and disulfide bonds (s=s) of WT proinsulin are shown. Amino acid substitutions result of insulin-induced hypoglycemia in both patients associated with PNDM/MDI are shown in yellow.

2150 The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 research article

Figure 3 Secondary structure computer modeling of WT and mutant proinsulins. Shown are superimposed Cα traces of WT (blue) and all mutated insulins (purple), with the exception of LB6V. The positions of disulfide bridges are also marked. None of the mutations caused substantial distortion of the secondary structure, with the exception of LB15YB16delinsH (yellow trace at left), where an α-helical disruption is apparent (arrow). This was also evident in the superimposition of WT insulin and LB15YB16delinsH. The LB6P mutation alters the hydrophobic core of the protein. The LB11P mutation affects the hydrophobic core of the protein. The CA6Y mutation disrupts the A6–A11 disulfide bridge; the tyrosine is oriented inside the hydrophobic core of the protein, where it engages LB6 in a stacking interaction. The hAkita mutation — used in this study as positive control — disrupts the B7–A7 disulfide bridge, and the tyrosine is solvent exposed.

occurred at 37–1,460 d postpartum (median, 88 ± 412.4 d), which — permanent or transient — with onset during the first years of is about 5 wk later than onset in patients with KCNJ11 mutations life caused by a single gene defect. (range, 1–220 d; median, 53 ± 58.3 d; P < 0.07). With exclusion of Functional studies. Figure 2 shows the location of the mutated resi- 1 case (kindred C, I-1, 1,460 d), the comparison reached statistical dues. Two mutations each affect 1 of the 3 highly conserved insulin significance (median, 87 ± 94.7 d; P < 0.00038). Indeed, individu- disulfide bonds, similar to that reported in the Akita mouse: CA6Y als with INS gene mutation can present with diabetes well beyond perturbs CA6–CA11, while YA19X disrupts CB19–CA20 by generating a 6 mo after birth (F. Barbetti, unpublished observations). This fea- premature stop codon that eliminates CA20 as well as NA21 at the C ture is also found in patients with EIF2AK3 mutations (Wolcott- terminus. Another mutation is an in-frame substitution in which Rallison syndrome), which disrupt the normal ER stress response B-chain Leu15 and Tyr16 are replaced by a single His residue: and cause dysregulated secretory protein synthesis in pancreatic LB15YB16delinsH. In 3 additional cases, proinsulin mutations were β cells (6, 14). Finally, because most INS gene mutations occur located in invariant residues of the B-chain (LB6P, LB6V, and LB11P; spontaneously, these patients cannot be distinguished on clini- Supplemental Figure 1). Of these, LB6P and LB11P are located in the cal grounds from those with type 1 diabetes (besides the absence so-called “hydrophobic core” on the contact surface buried within of type 1 diabetes autoantibodies). For this reason we propose insulin dimers. The mutations predicted to most dramatically affect here to replace the term PNDM with monogenic diabetes of infancy proinsulin secondary structure included LB15YB16delinsH (by caus- (MDI), a broad definition that may include any form of diabetes ing significant α-helix disruption; Figure 3) and R65C (by abolish-

Figure 4 Misfolding and defective secretion of proinsulin mutants. 293T cells were transfected with empty vector or cDNAs encoding WT proinsulin or the following proinsulin mutants: LB6P (PB6); LB11P (PB11); LB15YB16delinsH (HB15,16); YA19X (XA19); CA6Y (YA6); R65L (L65); R65C (C65); and hAkita. (A) Transfected cells were metaboli- cally labeled with 35S–amino acids for 1 h and then further chased for 1 h. Cell lysates (C) and chase media (M) were immunoprecipitated with anti-insulin, and the samples were analyzed by nonreducing Tris-tricine-urea- SDS-PAGE. All proinsulin (Pro) disulfide isomers dem- onstrated different mobilities; the native form is the fast- migrating, secreted form obtained for WT proinsulin. (B) Recombinant proinsulin secreted for 16 h into serum-free medium was quantified by RIA.

The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 2151 research article

Figure 5 Viability of transfected INS-1E cells. Shown is the viability of INS-1E cells at 48 and 96 h after transfection with WT human (pro)insulin, hAkita mutation, human familial hyper(pro)insulinemia mutation R65L, and human MDI mutations R65C, CA6Y, LB15YB16delinsH, LB6P, LB11P, and YA19X. Human (pro)insulin was detected by monoclonal antibody directed toward the human C-peptide and C terminus of the B-chain of the (pro)insulin molecule. Few (LB6P and LB11P) or no (hAkita, R65C, CA6Y, LB15YB16delinsH, and YA19X) INS-1E cells expressing MDI mutations were visible 96 h after transfection.

ing the C-peptide/A-chain cleavage site while creating an unpaired Next, WT proinsulin, the R65L mutant, the hAkita mutation, Cys residue with potential to form inappropriate disulfide bonds). and all new MDI-associated mutations except LB6V were expressed In 3 independent experiments, all proinsulin mutants except as recombinant human proinsulins in INS-1E rat insulinoma LB6V were expressed as recombinant proteins in HEK293 cells cells, which are capable of processing proinsulin into insulin and and examined for 35S-protein expression and secretion at 1 h after C-peptide. The recombinant proinsulins were readily detected by synthesis (Figure 4A). After immunoprecipitation from cells and immunofluorescence with a human-specific anti-proinsulin anti- analysis by nonreducing Tris-tricine-urea-SDS-PAGE, WT pro- body at 48 h after transfection (Figure 5). However, at 96 h after trans- insulin and proinsulin-R65L (a hyperproinsulinemia-inducing fection, immunofluorescence was still detectable in 61% ± 2.8% of mutation associated with mild fasting hyperglycemia in an adult INS-1E cells transfected with WT proinsulin (n = 4 transfections) and patient; ref. 19) were well expressed, with similar fractions of fast- 56% ± 4.0% of cells transfected with the R65L mutation (n = 3 trans- migrating native proinsulin disulfide isomer and slower-migrating fections), but almost none was detected in cells transfected with nonnative forms (34). Within 1 h of synthesis, both WT proinsulin the hAkita mutation or with R65C, LB15YB16delinsH, CA6Y, or YA19X and the R65L mutant underwent efficient secretion to the culture (mean, 3 transfections each). An intermediate percentage of viable media. However, all MDI-causing mutations and the Akita muta- INS-1E cells at 96 h was observed for mutations LB6P (36% ± 2.2%; tion introduced into human proinsulin (hAkita; mutation CA7Y) n = 2 transfections) and LB11P (17.7% ± 4.8%; n = 5 transfections). exhibited substantially decreased recovery with an increased frac- The same constructs tested by immunofluorescence were used tion of nonnative proinsulin disulfide isomers and a profound to evaluate XBP1 gene splicing, a marker of ER stress. Tunicamy- decrease in secretion (Figure 4A). In a separate set of experiments, cin (5 μg/ml for 4 h), a general inducer of ER stress, was used as proinsulin mutants that were not secreted appeared to be degrad- positive control. As seen in Figure 6A, no difference was detected ed intracellularly (data not shown). As an independent assay, pro- between untransfected cells and those transfected with WT proinsulin released into serum-free culture media was detected by RIA insulin or the R65L mutation. In contrast, recombinant human (Figure 4B), confirming defective secretion for all MDI-producing proinsulins containing MDI-associated mutations (including proinsulin mutants. These data establish that MDI-producing LB6P; data not shown) and the hAkita mutation induced increases proinsulin mutations perturb proinsulin structure and lead to of the shorter, splice-activated form of XBP1 from 30% (R65C defective secretion. and YA19X) to 400% (CA6Y) (Figure 6B). We then tested protein

2152 The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 research article

Figure 6 XBP1 splicing in transfected HEK293 cells. (A and B) Splicing of transcription factor XBP1 by agarose gel (A) and densitometry (B). Shown is the percentage of spliced/unspliced XBP1 forms for each sample, as assessed by densitometry, relative to WT. The red line indicates ratio 1. All mutations associated with PNDM or infancy-onset diabetes (and hAkita) induced an increase in XBP1 splicing compared with WT (pro)insulin or familial hyper(pro)insulinemia mutation R65L.

FB24S, VA3L; Insulin Chicago, Los Angeles, and Wakayama; refs. 27, 28, 30) or in the dibasic C-peptide/A-chain cleavage site (R65H, R65L, R65P; Proinsulin Tokyo and Kyoto; refs. 19–21, 31, 32), with the exception of the HB10D mutant (Proinsulin Providence) that may affect insulin packaging (29). However, in familial hyperproinsulinemia, the abnormal INS gene product is efficiently secreted. In addition, no INS coding sequence polymorphisms have been found associated with type 2 diabetes (35, 36). In this study we describe what we believe to be new heterozygous mutations within the proinsulin coding sequence causing early-onset, insulin-requiring diabetes (2–48 mo postpartum). In the Italian patient cohort with PNDM, mutations of the INS gene were the second leading cause (9 of 37, or 24.3%) following activating mutations of KCNJ11 (19 of 37, or 51.3%). Spontaneous mutations in a small expression of chaperonin Grp78, also known as BiP, as an addi- gene are unlikely to be a chance observation (9), yet in 7 patients tional marker of ER stress. As shown in Figure 7, Grp78 protein the mutations arose de novo. We also describe 2 familial cases with amounts increased in 293 cells 9 and 16 h after transfection with vertical transmission of insulin-requiring diabetes. In 2 patients CA6Y and LB15YB16delinsH mutant proinsulins compared with (kindred A, II-1, and kindred H) in which hyperglycemia was WT proinsulin; higher levels of Grp78 were observed at 24 h after detected before severe diabetes symptoms, C-peptide was found transfection for mutations LB6P and LB11P. at normal to high levels, but was decreased or undetecable within As an early apoptosis marker, we assessed surface staining of 2–12 mo of diagnosis. The INS gene mutations described herein annexin V. In 3 independent flow cytometry experiments at 24 h are highly reminiscent of the diabetes phenotype observed in male after transfection of HEK293 cells, we detected a 4-fold increase of Akita mice, in which a heterozygous mutation of the Ins2 gene annexin V positivity in cells expressing proinsulin LB15YB16delinsH (CA7Y, disrupting an interchain disulfide bond of insulin) causes a compared with cells expressing WT proinsulin (P = 0.015; Figure doubling of plasma glucose within 4 wk of birth (37) with a rapid 8A). However, using this technique, we were unable to observe decrease of pancreatic β cells (22). A proposed molecular patho- effects with other MDI mutations or even the hAkita mutation. genesis of diabetes in the Akita mutant is (a) proteotoxic proinsu- As a late apoptosis marker, we evaluated by flow cytometry the lin misfolding, (b) engorgement of the β cell ER with activation of percentage of propidium iodide–positive cells at 24, 48, and ER stress response pathways, and (c) generation of downstream 72 h after transfection (Figure 8B). With this assay, we detected increased numbers of apoptotic 293 cells transfected with the LB15YB16delinsH and the CA6Y mutations at 24 h after transfection compared with WT proinsulin (P < 0.003 and P < 0.01, respectively), but for the remaining MDI mutations we did not find any change at all points tested. Taken together, these data suggest that at least some of the proinsulin mutants, in addition to being defective in insulin production (loss of function), might also compro- mise β cell viability (toxic gain of function).

Discussion Previously, abnormal circulating proinsulin had been reported in subjects with elevated plasma levels of immunoreactive insulin but Figure 7 normal insulin sensitivity (19, 21, 27, 28, 30). Such patients with Grp78 expression. Western blot with anti-Grp78 primary antibody in HEK293 cells transfected with WT and mutated insulin. Cytoplasmic familial hyperproinsulinemia are diagnosed as adults with mild fractions were collected at 9, 16, and 24 h after transfection. Expres- diabetes, impaired glucose tolerance (19–21, 27, 28, 30, 32), or sion of β-tubulin was assessed on the same filters to normalize the even episodic hypoglycemia (26, 31). Mutations of the human INS amount of loaded proteins. Increased Grp78 protein expression was gene causing familial hyperinsulinemia or familial hyperproinsu- clearly visible at 9 and 16 h for mutations LB15YB16delinsH and CA6Y linemia encode defects either in insulin receptor binding (FB25L, and at 24 h for mutations LB6P and LB11P.

The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 2153 research article

Figure 8 Annexin V expression and propidium iodide assay. (A) Percentage of apoptotic cells, as assessed by annexin V expression, in HEK293 cells transfected with human WT (pro)insulin or LB15YB16delinsH mutation. Expression of mutant (pro)insulin effected a 3-fold increase in annexin V– positive cells 24 h after transfection compared with cells transfected with WT (pro)insulin. (B) Percentage of propidium iodide–positive cells transfected with WT, LB15YB16delinsH, and CA6Y mutation cDNA. P values are shown compared with WT (pro)insulin.

mutation presented with diabetic ketoacidosis at 86 and 134 d after birth. However, this result indicates that such rare patients with INS gene mutants are likely to be hidden within the larger population of those diagnosed with type 1 diabetes. Finally, we advocate the concept of substituting the term MDI for PNDM (41), reflecting our awareness that onset of hyperglycemia in monogenic forms of diabetes extends beyond the neonatal period (refs. 6, 9, 15, and the present study).

Methods Subjects. The Early Onset Diabetes Study Group of the Italian Society of Pediatric Endocrinology and Diabetes (SIEDP) has been studying 37 patients who had diabetes onset within the first 220 d of life and are negative for type 1 diabetes autoantibodies. Of these probands, we detected activating heterozygous mutations of KCNJ11 in 19 subjects (refs. 13, 16, 17, and C. Colombo and F. Barbetti, unpublished observations) and a homo- apoptotic signals (22–25). The β cell deficiency in male Akita mice zygous inactivating mutation of GCK in a single case (7). We screened the stands in striking contrast with the finding that Ins1–/–Ins2–/– mice, remaining 17 patients for mutations in the coding region of the INS gene which lack all insulin and die of neonatal diabetes, actually show and genotyped the parents of each proband with an INS mutation. The hyperplastic pancreatic islets (38). INS gene was also examined in 2 additional patients from Denmark with The translational studies described herein support the hypothe- infancy-onset diabetes. Written informed consent was obtained from all sis that all of the new INS gene mutations reported in patients with participants or their parents. The study was approved by the ethics com- PNDM/MDI cause proinsulin misfolding. Two mutants described mittee of the Early Onset Diabetes Study Group of the SIEDP. in this report, CA6Y and YA19X, directly lack 1 of the 3 stabilizing Clinical studies. We compared the clinical features of the patients carry- disulfide bonds of proinsulin (39), while the mutants LB6P, LB6V, ing mutations of the INS gene with those of patients from our case series LB11P, and LB15YB16delinsH involve less obvious structural distor- carrying a KCNJ11 mutation. An insulin tolerance test was performed by tions. The clinical phenotype of the patient carrying the YA19X injecting 0.1 IU/kg BW i.v. of regular insulin. Blood samples were with- mutation (diabetes onset at age 88 d) seems to indicate that the drawn at –15, 0, 15, 30, and 45 min relative to the time of insulin injection mutant protein is translated and that the mRNA decay usually for determination of plasma glucose, pancreatic glucagon (Linco), growth associated with a premature stop codon is not involved. Further hormone, and cortisol. evidence for proinsulin misfolding comes from an increased ER Functional studies. Insulin cDNA was kindly provided by H. Lou (NIH, stress response (Figure 6). Moreover, examination of R65C, found Bethesda, Maryland, USA). Full-length insulin cDNA was amplified by PCR in 3 independent probands, as well as YA19X and CA6Y, like the using the following primers containing HindIII or EcoRI restriction sites: Akita mutant (Figure 4), revealed the presence of improper disul- Ins-Hind, 5′-CCTAGTTGAGAAGCTTGTCCTTCTGCCATGGCCCTG-3′; fide bonds with little or no secretion, consistent with entrapment Ins-Eco, 5′-CCTAGTTGAGGAATTCTCCATCTCTCTCGGTGCAGG-3′. within the ER. The direct demonstration of cell death induced by Amplification product was digested with EcoRI and HindIII, run on 1% expression of at least 2 of the mutants (Figure 8) strongly supports agarose gel, purified using QIAquick Gel Extraction kit (Qiagen), and a proteotoxic, apoptotic mechanism. cloned into pCDNA3.1+ expression vector (Invitrogen). Exon 2 of insulin Of the patients described herein, 9 of 10 showed normal weight LB15YB16delinsH was directly cloned after PCR amplification in pCRII- at birth, a finding clearly different from the low birth weight in TOPO vector of the TOPO TA cloning system (Invitrogen). All newly patients bearing KCNJ11 mutations (15), which suggests that described mutations, with the exception of LB6V, as well as control muta- β cell insufficiency may occur primarily after birth. Pancreatic tions R65L and hAkita (22) were introduced by site-directed mutagenesis β cell proliferation is low in humans after birth (40). A postpartum of the human proinsulin cDNA using QuickChangeII Site directed muta- decline in measurable C-peptide, taken together with other results genesis kit (Stratagene) following the manufacturer’s protocol. presented herein, support the hypothesis that postnatal failure to INS-1E cells (gift of C. Wollheim, University Medical Centre, Geneva, maintain β cell mass due to proteotoxic proinsulin misfolding is Switzerland) were cultured in RPMI (Invitrogen) supplemented with a primary cause of PNDM in these patients with INS gene muta- 2 mM l-glutamine, 10 mM HEPES, 50 μM β-mercaptoethanol, 10% FBS, tions. We are not able at this stage of our study to explain why the 100 IU/ml penicillin, and 100 μg/ml streptomycin in a 5% CO2 atmo- parents of familial cases with the R65C mutation were diagnosed sphere at 37°C. INS-1E cells (80% confluent) were transfected transiently with diabetes at 1 and 4 yr of age, while 2 probands with the same using Lipofectamine (Invitrogen) according to the manufacturer’s proto-

2154 The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 research article col. Expression of WT proinsulin and each new proinsulin mutant (with tides (i.e., proinsulin) using an anti-rat insulin RIA (Millipore) that reacts the exception of LB6V) plus control proinsulin-R65L and hAkita mutants with WT and mutant proinsulins. were visualized at 48 and 96 h after transfection by immunofluorescence Apoptosis assays. HEK293 cells were seeded in a 24-well plate (150,000 cells/ using monoclonal antibody to human proinsulin (1:300 dilution; Abcam) well) 24 h before transfection. Cells at 50%–60% confluence were transiently and Alexa Fluor 594 goat anti-mouse (1:500 dilution; Invitrogen). Mono- transfected with 800 ng plasmid containing WT proinsulin, a proinsulin clonal antibody to human proinsulin has no cross-reactivity with either mutant (with the exception of LB6V), human familial hyper(pro)insulinemia rat proinsulin 1 or rat proinsulin 2. Transfected cells were visualized by R65L, or hAkita. To assay annexin V surface staining, adherent cells at 24 immunofluorescence after the following procedure: cells were washed and 48 h after transfection were recovered by trypsinization and collected with PBS, fixed in 4% paraformaldehyde for 10 min, washed quickly in with complete DMEM. Each culture medium was collected and added to PBS, permeabilized in PBS plus 0.2% Triton X-100 for 15 min, blocked in the corresponding sample in order to count detached cells. The annexin V PBS plus 10% FBS for 30 min, incubated with primary antibody for 1 h at assay (Calbiochem) was performed according to the manufacturer’s pro- 37°C, washed 3 times with PBS plus 0.5% Tween, incubated with second- tocol. The number of apoptotic cells was measured by FACSCalibur flow ary antibody for 1 h at 37°C, washed 3 times with PBS plus 0.5% Tween, cytometer (Becton Dickinson). In the propidium iodide assay, cells were and mounted with 50% glycerol. Insulin-expressing cells were visualized similarly recovered by trypsinization, fixed with 70% cold ethanol, treated with a Nikon TE2000-E Eclipse fluorescence microscope and counted with RNase, and stained with 20 ng/ml propidium iodide. The number of along 2 diameters of each well. apoptotic cells was measured by flow cytometry as described above. The XBP1 splicing. HEK293T cells were transfected with WT or mutant INS percentage of apoptosis in untransfected, control HEK293 cells (4%, mean cDNA. At 2 d after transfection, RNA was extracted and retrotranscribed, of 5 experiments) was subtracted from the percentage of apoptotic cells and 50 ng was used for PCR amplification of XBP1. Amplified fragments observed in transfections with WT or mutant (pro)insulins. were resolved by agarose gel electrophoresis and scanning densitometry. Modeling. Computer modeling of the structures of human WT and Grp78 protein expression. HEK293T cells’ cytoplasmic proteins were mutant insulins were generated by homology to bovine insulin as a template extracted in lysis buffer (containing 0.5% Triton X, 1 mM DTT, 1 mM using Modeller software version 8.1 (http://www.salilab.org/modeller/). PMSF, 4 mg/ml aprotinin, and 2 μg/ml pepstatin) at 9, 16, and 24 h after Ramachandran plots of the models indicated that there were no residues transfection with WT or mutant INS cDNA, after which they were loaded in disallowed regions. on SDS-PAGE gels at 10% and electroblotted. Filters were decorated with Statistics. Values are expressed as mean ± SEM. A 2-tailed Student’s anti-Grp78 antibody (catalog no. PA1-41051, 1:1,000 dilution; Affinity t test was used for all statistical analysis. P values less than 0.03 were Bioreagents) followed by a peroxidase-conjugated goat anti-rabbit sec- considered significant. ondary antibody for enhanced chemiluminescence. Anti-human β-tubulin (Santa Cruz Biotechnology Inc.) was used at 1:1,000 dilution with the Note added in proof. Since submission of this report, 4 papers have same secondary antibody. been published describing INS gene mutations associated with Biosynthesis of WT proinsuiln and mutations R65L, R65C, LB6P, LB11P, PNDM (42–45). LB15YB16delinsH, CA6Y, YA19X, and hAkita. Transfected HEK293T cells were preincubated in Met/Cys-deficient DMEM, pulse-labeled with 35S–amino Acknowledgments acids, and then chased in complete media. Transfected cells were lysed in We thank all the families who participated in this study. P. Arvan’s 0.1% SDS, 1% Triton X-100, 0.2% deoxycholate, 100 mM NaCl, 10 mM laboratory was supported by NIH grant R01-DK48280. EDTA, and 25 mM Tris, pH 7.4, plus a proteinase inhibitor cocktail. Lysed cells and chase media were subjected to immunoprecipitation with guinea Received for publication August 30, 2007, and accepted in revised pig anti-insulin, and samples were analyzed by Tris-tricine-urea-SDS-PAGE form March 19, 2008. as described previously (33). To quantify proinsulin secretion, at 40 h after transfection cells were washed and then further incubated for 16 h in Address correspondence to: Fabrizio Barbetti, Department of serum-free DMEM containing 25.5 mM glucose plus 0.2 % bovine serum Internal Medicine, University of Tor Vergata, Viale Oxford 81, albumin (RIA-grade). The incubation media were collected, and aliquots 00134 Rome, Italy. Phone: 39-06-20900672; Fax: 39-06-20900676; were taken to measure total immunoreactivity of insulin-containing pep- E-mail: [email protected].

1. Polak, M., and Shield, J. 2004. Neonatal and very-early- 7. Njolstad, P.R., et al. 2001. Neonatal diabetes melli- 13. Tonini, G., et al. 2006. Sulphonylurea treatment onset diabetes mellitus. Semin. Neonatol. 9:59–65. tus due to complete glucokinase deficiency. N. Engl. outweighs insulin therapy in short-term metabolic 2. Gloyn, A.L., et al. 2005. Relapsing diabetes can J. Med. 344:1588–1592. control of patients with permanent neonatal dia- result from moderately activating mutations in 8. Bennet, C.L., et al. 2001. The immune dysregula- betes mellitus due to activating mutations of the KCNJ11. Hum. Mol. Genet.14:925–934. tion, polyendocrinopathy, enteropathy, X-linked KCNJ11 (Kir6.2) gene. Diabetologia. 49:2210–2213. 3. Colombo, C., et al. 2005. Transient neonatal syndrome (IPEX) is caused by mutations of FOXP3. 14. Seneé, V., et al. 2004. Wolcott-Rallison syndrome. diabetes mellitus is associated with a recurrent Nat. Genet. 27:20–21. Clinical, genetic and functional study of EIF2AK3 (R201H) KCNJ11 (KIR6.2) mutation. Diabetologia. 9. Gloyn, A.L., et al. 2004. Activating mutations in the mutations and suggestion of genetic heterogeneity. 48:2439–2441. ATP-sensitive potassium channel subunit Kir6.2 Diabetes. 53:1876–1883. 4. Babenko, A.P., et al. 2006. Activating mutations gene are associated with permanent neonatal dia- 15. Hattersley, A.T., and Ashcroft, F.M. 2005. Activat- in the ABCC8 gene in neonatal diabetes mellitus. betes. N. Engl. J. Med. 350:1838–1849. ing mutations in Kir6.2 and neonatal diabetes. New N. Engl. J. Med. 355:456–466. 10. Sellick, G.S., et al. 2004. Mutations in PTF1A cause clinical syndromes, new scientific insights, and new 5. Stoffers, D.A., Zinkin, N.T., Stanojevic, V., Clarke, pancreatic and cerebellar agenesis. Nat. Genet. therapy. Diabetes. 54:2503–2513. W.L., and Habener, J.F. 1997. Pancreatic agenesis 36:1301–1305. 16. Massa, O., et al. 2005. KCNJ11 activating muta- attributable to a single nucleotide deletion in the 11. Senee, V., et al. 2006. Mutations in GLIS3 are tions in Italian patients with permanent neonatal human IPF1 gene coding sequence. Nat. Genet. responsible for a rare syndrome with neonatal dia- diabetes. Hum. Mutat. 25:22–27. 15:106–110. betes mellitus and congenital hypothyroidism. Nat. 17. Masia, R., et al. 2007. An ATP-binding mutation 6. Delepine, M., et al. 2000. EIF2AK3, encoding trans- Genet. 38:682–687. (G334D) in KCNJ11 is associated with a sulfonyl- lation initiation factor 2-α kinase 3, is mutated in 12. Pearson, E.R., et al. 2006. Switching from insulin to urea-insensitive form of developmental delay, epi- patients with Wolcott-Rallison syndrome. Nat. oral sulfonylureas in patients with diabetes due to lepsy, and neonatal diabetes. Diabetes. 56:328–336. Genet. 25:406–409. Kir6.2 mutations. N. Engl. J. Med. 355:467–477. 18. Taylor, S.I., et al. 1992. Mutations in the insulin

The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008 2155 research article

receptor gene. Endocr. Rev. 13:566–595. of an emerging new clinical syndrome. N. Engl. J. 37. Yoshioka, M., Kayo, T., Ikeda, T., and Koizumi, A. 19. Yano, H., et al. 1992. A novel point mutation Med. 310:1288–1294. 1997. A novel locus, Mody4, distal to D7Mit189 on in the human insulin gene giving rise to hyper 29. Gruppuso, P.A., et al. 1984. Familial hyperproinsu- chromosome 7 determines early-onset NIDDM in proinsulinemia (Proinsulin Kyoto). J. Clin. Invest. linemia due to a proposed defect in conversion of nonobese C57BL/6 (Akita) mutant mice. Diabetes. 89:1902–1907. proinsulin to insulin. N. Engl. J. Med. 311:629–634. 46:887–894. 20. Nakashima, N., Sakamoto, N., Umeda, F., et al. 30. Nanjo, K., et al. 1986. Diabetes due to secretion of a 38. Duvillie, B., et al. 1997. Phenotypic alterations in 1993. Point mutation in a family with hyperpro- structurally abnormal insulin (insulin Wakayama). insulin-deficient mutant mice. Proc. Natl. Acad. Sci. insulinemia detected by single stranded confor- Clinical and functional characteristics of [LeuA3] U. S. A. 94:5137–5140. mational polymorphism. J. Clin. Endocrinol. Metab. insulin. J. Clin. Invest. 77:514–519. 39. Rhodes, C.J. 2004. Processing of the insulin mol- 76:633–636. 31. Barbetti, F., et al. 1990. Two unrelated patients ecule. In Diabetes Mellitus. A fundamental and clinical 21. Oohashi, H., et al. 1993. Familial hyperproinsu- with familial hyperproinsulinemia due to a muta- text. 3rd edition. D. LeRoith, S.I. Taylor, and J.M. linemia associated with NIDDM. A case study. Dia- tion substituting histidine for arginine at position Olefsky, editors. Lippincott, Williams & Wilkins. betes Care. 16:1340–1346. 65 in the proinsulin molecule: identification of the Philadelphia, Pennsylvania, USA. 27–50. 22. Wang, J., et al. 1999. A mutation in the insulin 2 mutation by direct sequencing of genomic DNA 40. Kassem, S.A., Ariel, I., Thornton, P.S., Scheim- gene induces diabetes with severe pancreatic β-cell amplified by polymerase chain reaction. J. Clin. berg, I., Glaser, B. 2000. Beta-cell proliferation and dysfunction in the Mody mouse. J. Clin. Invest. Endocrinol. Metab. 71:164–169. apoptosis in the developing normal human pan- 103:27–37. 32. Warren-Perry, M.G., et al. 1997. A novel point muta- creas and in hyperinsulinism of infancy. Diabetes. 23. Zuber, C., Fan, J.-Y., Guhl, B., and Roth, J. 2004. tion in the insulin gene giving rise to hyperproinsu- 49:1325–1333. Misfolded proinsulin accumulates in expanded linemia. J. Clin. Endocrinol. Metab. 82:1629–1631. 41. Barbetti, F. 2007. Diagnosis of neonatal and infancy- pre-Golgi intermediates and endoplsmic reticulum 33. Shuldiner, A.R., Barbetti, F., Raben, R., Scavo, L., and onset diabetes. Endocr. Dev. 11:83–93. subdomains in pancreatic beta cell of Akita mice. Serrano, J. 1990. Insulin. In Insulin-like growth factors: 42. Støy, J., et al. 2007. Insulin gene mutations as a FASEB J. 18:917–919. molecular and cellular aspects. D. LeRoith, editor. CRC cause of permanent neonatal diabetes. Proc. Natl. 24. Oyadomari, S., et al. 2002. Targeted disruption of Press. Boca Raton, Florida, USA. 181–219. Acad. Sci. U. S. A. 104:15040–15044. the Chop gene delays endoplasmic reticulum stress- 34. Liu, M., Li, Y., Cavener, D., and Arvan, P. 2005. 43. Edghill, E.L., et al. 2008. Insulin mutation screen- mediated diabetes. J. Clin. Invest. 109:525–532. Proinsulin disulfide maturation and misfolding in 1,044 patients with diabetes: mutations 25. Ron, D. 2002. Proteotoxicity in the endoplasmic ing in the endoplasmic reticulum. J. Biol. Chem. in the INS gene are a common cause of neonatal reticulum: lessons from the Akita diabetic mouse. 280:13209–13212. diabetes but a rare cause of diabetes diagnosed in J. Clin. Invest. 109:443–445. 35. Raben, N., et al. 1991. Normal coding sequence of childhood or adulthood. Diabetes. 57:1034–1042. 26. Gabbay, K.H., DeLuca, K., Fisher, J.N., Mako, M.E., insulin gene in Pima Indians and Nauruans, two 44. Polak, M., et al. 2008. Heterozygous missense muta- and Rubenstein, A.H. 1976. Familial hyperproinsu- groups with highest prevalence of type II diabetes. tions in the insulin gene are linked to permanent linemia. An autosomal dominant defect. N. Engl. J. Diabetes. 40:118–122. diabetes appearing in the neonatal period or in early Med. 294:911–915. 36. Olansky, L., Janssen, R., Welling, C., and Permutt, infancy: a report from the French ND (Neonatal 27. Given, B.D., et al. 1980. Diabetes due to secretion of M.A. 1992. Variability of the insulin gene in Diabetes) Study Group. Diabetes. 57:1115–1119. an abnormal insulin. N. Engl. J. Med. 302:129–135. American blacks with NIDDM. Analysis by single- 45. Molven, A., et al. 2008. Mutations in the insulin 28. Haneda, M., et al. 1984. Familial hyperinsulinemia strand conformational polymorphisms. Diabetes. gene can cause MODY and autoantibody-negative due to a structurally abnormal insulin. Definition 41:742–749. type 1 diabetes. Diabetes. 57:1131–1135.

2156 The Journal of Clinical Investigation http://www.jci.org Volume 118 Number 6 June 2008